Rackspace blog post (#1819)
## Description Adds a slightly-edited version of the Rackspace blog post.
This commit is contained in:
parent
a9123d4080
commit
db415c1ffb
237
website/blog/2022-12-16-not-your-backups-not-your-data.md
Normal file
237
website/blog/2022-12-16-not-your-backups-not-your-data.md
Normal file
@ -0,0 +1,237 @@
|
|||||||
|
---
|
||||||
|
slug: not-your-backups-not-your-data
|
||||||
|
title: "Not your backups, not your data: A Rackspace story"
|
||||||
|
description: "Rackspace email outage: What happened, what went wrong, best practices, and paths forawrd to protect your business email and data."
|
||||||
|
authors: nica
|
||||||
|
tags: [corso, microsoft 365, ransomware]
|
||||||
|
date: 2022-12-16
|
||||||
|
image: ./images/not_your_backups_not_your_data.png
|
||||||
|
---
|
||||||
|
|
||||||
|

|
||||||
|
|
||||||
|
Rackspace is having a rough few days.
|
||||||
|
|
||||||
|
## The story thus far
|
||||||
|
|
||||||
|
Rackspace Technology announced on December 7th that, six days after
|
||||||
|
suffering a ransomware attack, they're still unable to determine when
|
||||||
|
normal email service will be restored to the thousands of hosted
|
||||||
|
Microsoft Exchange customers. They asked for customers' patience and
|
||||||
|
provided temporary email fixes as they continue to investigate the
|
||||||
|
attack that began late last week and has caused ongoing disruptions.
|
||||||
|
|
||||||
|
Service interruptions started on December 5th with no end in sight.
|
||||||
|
|
||||||
|
<!-- truncate -->
|
||||||
|
|
||||||
|
“While the investigation is ongoing and in its early stages, at this
|
||||||
|
time, we're unable to provide any timeline or expectations for
|
||||||
|
restoration to the Hosted Exchange environment,” said an official
|
||||||
|
forum post last Wednesday, “we're working to provide customers with
|
||||||
|
archives of inboxes where available.”
|
||||||
|
|
||||||
|
Rackspace, which has been stressing many administrators out with its
|
||||||
|
lack of transparency and accessibility during the outage crisis, is
|
||||||
|
doing all it can to fix the disrupted service, including throwing as
|
||||||
|
many employees at the problem as possible. However, they remain
|
||||||
|
tight-lipped about what customer info might have been accessed.
|
||||||
|
|
||||||
|
## No end in sight
|
||||||
|
|
||||||
|
Rackspace has confirmed the attack was ransomware, and in the meantime
|
||||||
|
it’s recommending users shift to Microsoft 365, saying: “we're
|
||||||
|
working diligently to meet our customers' needs regarding access to
|
||||||
|
email and email forwarding. As noted in our recent updates, we've
|
||||||
|
arranged for all Hosted Exchange customers to have access to Microsoft
|
||||||
|
365”
|
||||||
|
|
||||||
|
[Read the full post from Rackspace
|
||||||
|
here.](https://status.apps.rackspace.com/index/viewincidents?service=6&start=1670216400)
|
||||||
|
|
||||||
|
The company doesn’t want to release full details of the incident,
|
||||||
|
saying that it’s under active investigation, but it can be
|
||||||
|
difficult to feel secure not knowing any details of the incursion.
|
||||||
|
|
||||||
|
## What went wrong
|
||||||
|
|
||||||
|
Rackspace is a large company, with 6,000 employees and over 3 billion
|
||||||
|
in annual revenue. As such this situation is a surprising one. The
|
||||||
|
company isn't sharing the details of the breach, but the issue isn't
|
||||||
|
that their security was breached, that production data stores were
|
||||||
|
compromised, or that this affected users' access. The issue is that
|
||||||
|
Rackspace has no plan for recovery (as far as customers are concerned,
|
||||||
|
an inadequate plan, as seen in [the Atlassian data
|
||||||
|
outage](https://www.atlassian.com/engineering/post-incident-review-april-2022-outage),
|
||||||
|
is the same thing as no plan). Demanding perfection isn't a path to
|
||||||
|
resilience.
|
||||||
|
|
||||||
|
Rackspace has been around for a long time and effective operations are
|
||||||
|
really their main selling point, as such they should have had an
|
||||||
|
effective:
|
||||||
|
|
||||||
|
- Backup system Disaster Recovery Plan Ransomware Plan
|
||||||
|
|
||||||
|
Notably, these three plans aren't identical, but they should all be in
|
||||||
|
place to ensure that the company can recover from any incident.
|
||||||
|
|
||||||
|
Backup systems are essential for ensuring data is safe and secure, as
|
||||||
|
well as providing a way to restore lost or corrupted files. Disaster
|
||||||
|
Recovery Plans provide guidance on how to respond when an unexpected
|
||||||
|
event occurs, such as a natural disaster or cyberattack. Ransomware
|
||||||
|
plans help protect against malicious actors who may attempt to hold
|
||||||
|
your data hostage by encrypting it until you pay them money.
|
||||||
|
|
||||||
|
As the saying goes, a backup is only as good as your recovery.
|
||||||
|
We believe that Rackspace likely had runbooks and backup and recovery
|
||||||
|
plans for the above but, as seen in the [Gitlab data loss
|
||||||
|
incident](https://about.gitlab.com/blog/2017/02/10/postmortem-of-database-outage-of-january-31/)
|
||||||
|
a few years ago, without frequent testing and verification, these
|
||||||
|
plans aren’t meaningful. These plans need to be updated to reflect the
|
||||||
|
constant security attacks we find ourselves under these days. Backups,
|
||||||
|
in particular, need to be hardened and made immutable to prevent
|
||||||
|
ransomware from modifying or deleting them.
|
||||||
|
|
||||||
|
It should also be stressed that a backup is meaningless if it can't
|
||||||
|
be restored in a short period of time. If an organization needs
|
||||||
|
multiple weeks to restore something as critical to a business as email
|
||||||
|
and data access, that's equivalent to a total data loss and can be a
|
||||||
|
company-ending event for the organization’s customers.
|
||||||
|
|
||||||
|
Rackspace needs to take steps immediately towards verifying and
|
||||||
|
testing these plans and implementing them into their operations so
|
||||||
|
that they can better prepare themselves for future incidents like this
|
||||||
|
one. They need to invest in better security measures such as
|
||||||
|
firewalls, antivirus software, intrusion detection systems (IDS), and
|
||||||
|
other tools designed specifically with ransomware protection in mind;
|
||||||
|
create detailed backup procedures; develop comprehensive recovery
|
||||||
|
strategies; train staff on proper response protocols; and regularly
|
||||||
|
test their system's resilience against potential threats.
|
||||||
|
|
||||||
|
That said, beyond pointing fingers, what can all of us learn from this?
|
||||||
|
|
||||||
|
## Do you *really* control your data?
|
||||||
|
|
||||||
|
This question, whether your data is really yours, always prompts
|
||||||
|
certain members of the chattering classes to say ‘this is why the
|
||||||
|
cloud was a mistake’ and recommend some kind of retrograde abandonment
|
||||||
|
of current technology. Such types won’t be happy until the first year
|
||||||
|
of every startup’s life is spent waiting for backordered RAM. To be
|
||||||
|
clear: we'ren’t recommending that you abandon cloud data storage and
|
||||||
|
hosting services.
|
||||||
|
|
||||||
|
However, key data ownership principles do dictate that you can't
|
||||||
|
fully trust any single public cloud tool with everything.
|
||||||
|
|
||||||
|
## “Not your keys, not your coins” ➡️ “Not your backups, not your data”
|
||||||
|
|
||||||
|
The phrase “Not your keys, not your coins” is a popular saying in the
|
||||||
|
cryptocurrency world. It's a reminder to users that if they don't
|
||||||
|
have control of their private keys, they don't have control of their
|
||||||
|
coins. Private keys are a string of numbers and letters that are used
|
||||||
|
to access a cryptocurrency wallet. They're the only way to access the
|
||||||
|
funds stored in the wallet, and if someone else has access to the
|
||||||
|
private keys, they can access the funds.
|
||||||
|
|
||||||
|
This principle applies to cloud data storage as well: if you have no
|
||||||
|
way of exporting all your data, and no backups independent from your
|
||||||
|
cloud provider, you don’t really ‘own’ your data and will be at the
|
||||||
|
mercy of whatever vagaries beset that provider.
|
||||||
|
|
||||||
|
Incidents like this one with Rackspace aren’t all that common, and
|
||||||
|
they shouldn’t be, but they do happen. When they do, you don’t want to
|
||||||
|
be at the mercy of the primary provider’s backup and restore SLAs (or
|
||||||
|
lack thereof). Instead, we recommend exploring some straightforward
|
||||||
|
ways to protect yourself by owning your own backups. If you control
|
||||||
|
your own backups, you also control where you can restore it for
|
||||||
|
temporary access and thereby sidestepping the failed infrastructure,
|
||||||
|
the latency of restores to prevent being at the back of the line in
|
||||||
|
the restore queue, the ability to permanently move your data to a
|
||||||
|
different infrastructure provider to reduce lock-in with unreliable
|
||||||
|
vendors, and more!
|
||||||
|
|
||||||
|
## Our recommendations
|
||||||
|
|
||||||
|
### Understand your provider’s responsibility model
|
||||||
|
|
||||||
|
As the data owner or IT lead, you should:
|
||||||
|
|
||||||
|
1. Research the cloud provider's Service Level Agreement (SLA) to
|
||||||
|
understand the responsibilities and liabilities associated with
|
||||||
|
service delivery (for example, see [Microsoft’s shared responsibility
|
||||||
|
model](https://learn.microsoft.com/en-us/azure/security/fundamentals/shared-responsibility)).
|
||||||
|
2. If your cloud or service provider doesn’t have a well-defined risk
|
||||||
|
model and you don’t have other reasonable options, ask as many
|
||||||
|
questions as possible during the on-boarding process in order to gain
|
||||||
|
a full understanding of your responsibilities and any other
|
||||||
|
obligations that need to be met. 3. Familiarize yourself with not
|
||||||
|
just your company’s backup and retention policies but also industry
|
||||||
|
standards and regulations related to data security, privacy, and
|
||||||
|
compliance with applicable laws that may apply when utilizing cloud
|
||||||
|
services from your particular provider.
|
||||||
|
|
||||||
|
### Have you own backups on an independent platform
|
||||||
|
|
||||||
|
While we disagree with the [3–2–1 backup
|
||||||
|
rule](https://www.uschamber.com/co/run/technology/3-2-1-backup-rule)
|
||||||
|
in today’s cloud environments (more in a different blog post), you
|
||||||
|
shouldn't store your backups on the same system or infrastructure as
|
||||||
|
your primary data. Otherwise, you will likely lose access, often
|
||||||
|
permanently, to your backups when you need them the most.
|
||||||
|
|
||||||
|
However, you also don’t need to buy on-prem backup appliances, build a
|
||||||
|
bespoke storage system, or jump through any hoops to create an
|
||||||
|
independent backup copy. Find a modern backup tool that optimizes for
|
||||||
|
usability and supports cloud object storage (S3, Azure Blob,
|
||||||
|
etc.). These tools should support deduplication, compression, and
|
||||||
|
encryption to reduce network and storage costs and you should pick a
|
||||||
|
cold storage tier to even further reduce any associated backup
|
||||||
|
expenses.
|
||||||
|
|
||||||
|
### Test and verify
|
||||||
|
|
||||||
|
Taking backups isn’t enough, you’ve got to try using them, even if
|
||||||
|
only as a test. On a frequent basis, ensure that your backups are
|
||||||
|
definitely encrypted with keys that you control, use your backup tool
|
||||||
|
can verify that the data hasn’t been corrupted (using cryptographic
|
||||||
|
hashing), and perform test restores into sandbox
|
||||||
|
environments. Ideally, this should be automated with alerts in case of
|
||||||
|
failures.
|
||||||
|
|
||||||
|
On a semi-frequent basis, do test restores of varying subsets of your
|
||||||
|
data and measure how long it takes to restore it, where the
|
||||||
|
bottlenecks might be, and to see how long it would take to recover a
|
||||||
|
meaningful amount of data from a major outage. Does that meet your
|
||||||
|
SLA? Note that this often doesn’t mean restoring all data immediately
|
||||||
|
but a subset of the most recently accessed and modified data to get
|
||||||
|
things back up and running.
|
||||||
|
|
||||||
|
## Backing up Microsoft 365
|
||||||
|
|
||||||
|
We would be remiss if we didn’t concretely tie our recommendations
|
||||||
|
back to where we started, a large email outage at Rackspace. Rackspace
|
||||||
|
has recommended that all customers move to Microsoft 365 and so, we
|
||||||
|
will focus on that. Microsoft’s architecture should be a lot more
|
||||||
|
stable but the customer’s responsibility hasn’t gone away and
|
||||||
|
Microsoft [
|
||||||
|
highlights](https://learn.microsoft.com/en-us/azure/security/fundamentals/shared-responsibility)
|
||||||
|
that on their website when they say that they're responsible for the
|
||||||
|
infrastructure but the customer is responsible for their data.
|
||||||
|
|
||||||
|
To help with your portion of the responsibility, we recently release
|
||||||
|
[Corso](/), a free, secure, and open-source tool that protects your
|
||||||
|
Microsoft 365 data, at low cost, by securely and efficiently backing
|
||||||
|
up all business-critical data to object storage.
|
||||||
|
|
||||||
|
You can get started with Corso in [just a few
|
||||||
|
minutes](https://corsobackup.io/docs/quickstart) and you will be able
|
||||||
|
to move your Microsoft 365 emails and data to your choice of object
|
||||||
|
storage provider and location. Not only will Corso compress and
|
||||||
|
deduplicated your data, it will also encrypt it with your keys. Corso
|
||||||
|
also uses internal checksums to verify the integrity of your data.
|
||||||
|
|
||||||
|
It’s also possible to quickly verify and test your backups with
|
||||||
|
Corso. You can restore user data in sandbox environments and accounts
|
||||||
|
and perform filtered restores to just bring back subsets of data. Give
|
||||||
|
it a try and [let us know what you
|
||||||
|
think](https://discord.gg/63DTTSnuhT)!
|
||||||
BIN
website/blog/images/not_your_backups_not_your_data.png
Normal file
BIN
website/blog/images/not_your_backups_not_your_data.png
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 1.1 MiB |
@ -26,3 +26,12 @@ deduplicate
|
|||||||
vCPUs
|
vCPUs
|
||||||
cybercrime
|
cybercrime
|
||||||
cyberattacks
|
cyberattacks
|
||||||
|
Rackspace
|
||||||
|
AWS_SHARED_CREDENTIALS_FILE
|
||||||
|
cryptocurrency
|
||||||
|
backordered
|
||||||
|
Gitlab
|
||||||
|
cyberattack
|
||||||
|
Atlassian
|
||||||
|
SLAs
|
||||||
|
runbooks
|
||||||
Loading…
x
Reference in New Issue
Block a user